Artificial intelligence (AI) and machine learning (ML) technologies have transformed the way we interact with money and financial services companies. A common misconception is that both AI and ML are limited to numerical and written data, when they can be applied to video data too – known as computer vision.
Computer vision takes video data and applies AI and ML to it in order to extract actionable insights – insights that previously would have been identified by a human.
Use cases in financial services
There are various use cases for computer vision in the financial sphere delivering enhanced services and creating competitive advantage in the marketplace.
In the insurance sector, for example, insurers can expedite claims processing using visual data such as photos from a car accident to go along with the description provided during a customer’s phone call. Analysing the photos for damage to the vehicle helps insurers identify fraudulent claims, calculate the cost of repairing the damage and expediting the claim payment to the customer.
Retail banking can also benefit from computer vision as the technology can be used to provide banks with the next level of understanding of their customer’s intentions, before they have even spoken to an employee, simply by the way they behave. Ranging from interactions to natural language processing, computer vision enables the banking personnel to be trained in a way that best suits the customer’s needs.
These tools that recognise customers’ behaviour can also be applied to security, allowing banks to identify fraud. An example of how computer vision can be utilised by financial services is provided by Dell’s director, market and strategy, Wayne Arvidson who explains the same tools for identifying good customer behaviours can be used for identifying fraudulent actions.
“The beauty of it is, I can redeploy a retrained model after we have identified fraudulent activity in real-time,” he said. “Now the model is more refined and is looking for that type of behaviour, it will provide us with more insights and help identify those who are trying to circumvent the system.”
An evolved eye for spotting bad actors
One of the biggest problems a company can face in the financial sector, be it in regards to security, customer interactions, or payments, is human error. Originally, in a banking facility, you would have someone in a control room watching a screen for hours on end, making observations and decisions as they saw fit. However, they could easily make a mistake, potentially leading to something very costly to the bank. Computer vision removes the human liability and acts as an evolved version of CCTV – one that provides feedback on what is seen.
An added addition of computer vision is that financial organisations can relocate resources to another sector of the company – potentially a customer-facing one, providing a more engaging experience for those who visit in a branch. Arvidson adds: “As opposed to having an employee manually generate an insight and act on it, computer vision radically reduces response time and improves the level of service that a company is able to provide from a safety and security perspective.”
A question of privacy
Outdated CCTV tech that took a picture every half-second or so, creating a stop motion like video, has now been replaced by the demand for high resolution (even 4K) video quality. This increase in quality comes at an expense though, as it requires much more data which in turn is a lot more demanding on any infrastructure in place. Therefore, transmitting this back to a control room for someone to analyse is very taxing on resources.
With NVIDIA technologies, Dell offers a solution called Real Time Federated Analytics and Learning, which captures and inferences all the high-quality data, but only sends back the important results to the central location. This allows improvements to the model to be made in real time, while also making the data transfer much more manageable and streamlined. It is then fed back into the model which allows it to send even further insights, creating a loop of improvement.
“Real Time Federated Analytics and Learning platform is the only one with a zero-trust solution in the market,” said Arvidson. “It is called zero trust for two reasons. The first is that every bit of information is encrypted, but the second, is that the data isn’t sent anywhere. The data is ingested and acted on, but the only thing we’re transmitting is the results of the inferencing, not the data itself.
“This means there’s no personally identifiable information going back to the company, which means if someone were to intercept it and try and make alterations, it would all be gibberish because it’s just math information related to the inferencing and model.”
Coopetition is a huge buzzword in the fintech space currently, and it can be applied here. As customers cannot be identified, companies do not need to fear losing their clientele by sharing this information. However, there is an element of cooperation as information identifying bad actors can be shared between companies, which in turn benefits everyone.
Companies using Dell solutions in this way are only able to do so as a result of its NVIDIA partnership. As Real Time Federated Analytics and Learning deals with video data, the ability to extract the necessary data comes from the capabilities provided by the GPU (graphic processing unit) technology.
NVIDIA’s team helps partners through developer support and providing resources to get 10 times the efficiency out of the GPU tech, in turn, not only creating better insights for Dell to work with, but also making the price of implementing the technology cheaper as each NVIDIA GPU can do so much more.
Learn more about computer vision solutions from Dell Technologies and NVIDIA.